APPLIED PHYSICS LETTERS 92, 011921 2008

Melt dispersion versus diffusive oxidation mechanism for aluminum nanoparticles: Critical experiments and controlling parameters
Valery I. Levitas,a Michelle L. Pantoya, and Birce Dikici
Department of Mechanical Engineering, Texas Tech University, Lubbock, Texas 79409, USA

Received 13 September 2007; accepted 21 November 2007; published online 8 January 2008 Critical experiments were performed on Al and MoO3 thermites. The diameter and alumina shell thickness of the Al nanoparticles were varied, and ame propagation velocities were measured. The results strongly support the melt-dispersion mechanism and contradict the diffusion oxidation mechanism. The parameters that control the oxidation rate and ame velocity are justied and directions for the synthesis of Al nanoparticles which are opposite to the current directions based on diffusion oxidation are suggested. An equation for the ame velocity versus Al nanoparticle geometrical parameters, thermomechanical properties, and synthesis parameters is formulated. 2008 American Institute of Physics. DOI: 10.1063/1.2824392 When the diameter of the Al particle reduces to 20 120 nm in contrast to the traditional 1 100 m size , its reactivity drastically increases: ame propagation velocities V reach 1 km/ s for reactive nanocomposite thermites as opposed to 1 m / s for microcomposite thermites,1,2 and ignition delay times are reduced by three orders of magnitude.3 Knowledge of the oxidation mechanism for Al nanoparticles is of great importance since it impacts the direction of development in nanoparticle synthesis. There are currently two candidates. The rst one is the oxidation mechanism by diffusion of Al and oxygen atoms toward each other through the growing oxide shell.47 For relatively slow heating rates 103 K / s , this diffusion oxidation mechanism DOM is consistent with transmission electron microscopy studies and corresponds well with theoretical modeling.5 The oxidation time in the experiments in Ref. 4 exceeds 1 s. However, the oxidation time for experiments reported in Refs. 1 and 2 in which the heating rate is estimated at 108 K / s is ve orders of magnitude shorter and, consequently, oxidation cannot occur by DOM.8,9 To resolve this and some other contradictions between experiments and the DOM, a mechanochemical mechanism for fast oxidation for high heating rates called the melt-dispersion mechanism MDM has recently been suggested.8,9 It is shown, theoretically, that at fast heating rates the alumina shell for nanoparticles does not break due to the difference in thermal expansion coefcients of Al and alumina until the Al melts. The fast melting of Al is accompanied by a 6% volume expansion which creates high dynamic pressure 1 3 GPa in the molten Al. This high pressure overloads the alumina shell with hoop stresses that exceed the ideal theoretical strength of alumina th and cause its dynamic spallation. Immediately after shell spallation, pressure within the molten particle remains unchanged while at the bare Al surface it is 10 MPa due to gas pressure and surface tension.9 Subsequently, an unloading wave propagates to the center of the particle and creates a tensile pressure of the same order of magnitude. This tensile pressure exceeds the cavitation limit strength of liquid Al, and disperses the molten Al into small bare clusters that y at high velocity 100 250 m / s . The oxidation of these clusters is not limited by diffusion.
a

The goal of this letter is to perform critical experiments that conrm predictions based on the MDM theory. These experimental conrmations of the MDM enable the justication of parameters that control the rate of oxidation and V, and suggest unexpected directions of synthesis for Al nanoparticles. Traditional wisdom based on the diffusive mechanism of oxidation suggests that the ame rate will grow if a the radius of the Al particle R is reduced, since the ratio of surface area to volume increases and the particles burn time is proportional to R1.16 Ref. 5 ; b the thickness of the oxide shell is reduced, since this reduces the length of the diffusion path, the primary barrier for diffusive reaction; and c the Al content in a particle increases. The alumina shell does not participate in the reaction and is considered deadweight. The MDM predicts the opposite effects. i The radius of the Al particle and the oxide shell thickness are not, separately, controlling parameters. ii The key controlling parameter is the ratio of Al core radius R to shell thickness , M = R / . Note that Al content is an increasing function of M. Below the critical value M = 19, V reaches its maximum possible value and is independent of M. For M 19, the ame velocity grows with reduction in M and consequently Al content . iii To increase or decrease V for M 19 and xed Al core radius, the oxide shell thickness must be increased or decreased . Experiments to elucidate the operative mechanism i.e., MDM or DOM have been performed for Al particles with various R, , and M see Table I, where V and S are the averaged ame rate and standard deviation . Experiments not
TABLE I. Particle and velocity data. D nm 44 50* 80 110* 110 121 120* V m/s 950 789 948 721 810 765 947 S m/s 46 72 37 30 37 23 64

nm 2 1 2 1.5 2 2 4

M 10.00 24.00 19.00 35.67 26.50 29.25 14.00

Electronic mail: valery.levitas@ttu.edu.

0003-6951/2008/92 1 /011921/3/$23.00 92, 011921-1 2008 American Institute of Physics Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

011921-2

Levitas, Pantoya, and Dikici

Appl. Phys. Lett. 92, 011921 2008

d
FIG. 1. Flame speed vs Al particle core radius for several oxide thicknesses. It is evident that there is no relationship between ame speed and particle size. The triangles correspond to current data, while the squares represent data taken from Ref. 1.

marked with * were reported earlier in Ref. 1. The main conclusions in Ref. 1 were that below a certain particle diameter D = 2 R + = 80 nm, V reaches a maximum value and does not change. These results cannot be explained by the DOM, for which the burn time is a function of R1.16 Ref. 5 and the reason for the ame velocity independence of the particle size was not clear. However, in Ref. 1 the oxide shell thickness was constant. Three additional experiments in Table I have been chosen with R, , and M values selected such that unambiguous conclusions regarding the main controlling parameters could be made. All experiments were performed with loose powders low density mixtures of Al and MoO3, at 7%8% theoretical maximum density in a conned burning chamber. Visible emission data were collected at 51 000 frames/ s using a high speed camera; pressure measurements were also recorded using picosecond response time pressure transducers at 40 kHz see Ref. 1 for details . The results are reported for an optimal equivalence ratio that provides the maximum V for each particle reported in Table I. The experimental data are plotted as a function of R and M in Figs. 1 and 2, respectively. The following conclusions are made: a In agreement with item i , V is not a function of particle radius alone Fig. 1 ; it is different for the same R and different shell thicknesses. The main parameter controlling V is M = R / Fig. 2 . For M 19, the ame velocity reaches its maximum possible value and is independent of M, R, , and Al

content. For M 19, V grows with reduction in M. This is in agreement with item ii . For the Al particles with D = 120 nm and two different 2 and 4 nm , V for particles with smaller and consequently larger Al content was 765 m / s while for the maximum V was particles with the thicker 947 m / s. This is consistent with prediction iii based on the MDM since M = 33 for = 2 nm and M = 15 19 for = 4 nm but contradicts items b and c typical of the diffusive mechanism. For Al particles with D = 50 nm and = 1 nm, V is expected to be higher than for all other particles with larger D and according to diffusive mechanisms a c . However, the maximum V was only 785 m / s, since M = 24 19. Similarly, particles with D = 110 nm and = 1.5 nm should produce higher V than particles D = 120 nm and = 2 nm and = 4 nm according to diffusive mechanism a c . However, according to the MDM, V is lower because of the larger M. Based on previous experiments, it was concluded that V reaches its maximum value for particles with D 80 nm. Current experiments show that the maximum V was measured for Al particles with D = 120 nm and thick 4 nm oxide shells. Thus, V reaches a maximum value for M 19, independent of R and . The experimental data in Fig. 2 are in close agreement with the theoretical prediction of the volume fraction of melt f necessary to fracture of the oxide shell, versus M. This strongly supports the MDM and suggests that f is an integral physical controlling parameter. Namely, V is proportional to the volume fraction of molten Al that induces fracture of the oxide shell and participates in the MDM during time scales of 10 s i.e., while the ame front passes through . The remaining nondispersed Al may react over longer time scales but does not contribute to V. This correlation also suggests that, in the given range of parameters, cavitation of the molten Al core is not the limiting process since the entire molten Al core contributes to the reaction and ame velocity. Another conclusion is that the ultimate tensile strength of the alumina u reaches the estimated theoretical strength th = 11.33 GPa.9

The value f can be found from the fracture criterion that hoop stresses in the shell reach the ultimate tensile strength of the alumina u.9 Taking an expression for the hoop stress from Ref. 9, the fracture criterion can be transformed to Af 2 + Bf + C = 0, where m = 1 + 1 / M, A=6
m

K 2 + m3 MG2K2 ,

FIG. 2. Flame speed divided by Vmax = 950 m / s vs relative Al core radius M = R / . For various oxide shell strengths shown near the curves the lines K T G 2K 2 + 6 2 + m 3 M mK S + correspond to the volume fraction of melt f necessary to fracture of the KM 4G2 + 3m3K2 u , oxide shell; they are based on Eqs. 1 and 2 . Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

B = Km2

1 + M pg + 2

2G2 3K2

011921-3

Levitas, Pantoya, and Dikici

Appl. Phys. Lett. 92, 011921 2008

1G 2K 2

C = 6 2 + m3 M +m
2

KS TG2K2 4 2 + m3
2

1 + M pg + 2
u

2KSG2 3 KS + 2G2 K2

3KSK2m3 + 4G2 Ks + m3 1 K2

. In this equation, 3 m is the volumetric expansion during the melting of Al, K is the difference in bulk moduli between liquid and solid Al, KS is the bulk modulus of solid Al, G2 and K2 are the shear and bulk moduli of the oxide shell, pg is the pressure of the gas, 1 and 2 are the surface tensions at the aluminum-alumina interface and alumina-gas interface that appears during the reaction, respectively, T is the difference between the melting temperature and the passivation temperature T0, at which the initial oxide shell was formed, is the difference in linear thermal expansion coefand cients between solid Al and alumina. The ame velocity can be expressed in the form V = Vmax f = Vmax B + B2 4AC /2A, V = Vmax, for f = 1, for 0 f 1; 2

where Vmax is the maximum velocity that can be achieved in the experimental setup under study. All parameters in Eq. 1 can be found in Ref. 9; T0 = 300 K. While currently Eq. 1 is conrmed for a relatively narrow set of parameters Fig. 2 and for specic experimental conditions conned with loose powder , this equation may be predictive for a much wider range of parameters for which the MDM is operative. The dependence of f on M, T0, , and u was analyzed numerically in Ref. 9 for given material properties. The current study presents an explicit analytical expression and empirically connects f with V. Note that the dependences of f and V on are very weak for the considered experiments. Plots of f versus M for several values of u are presented in Fig. 2 and the signicance of the shells high strength is evident. The current trends in Al nanoparticle synthesis that are based on DOM are directed to decrease the particle size and the oxide shell thickness to keep the Al content high. The current results suggest the opposite direction should be taken in order to increase the reaction and ame velocities. The oxide shell thickness should be increased up to a maximum value for which the shell still possess the maximum strength due to a lack of defects, and to increase the particle size up to the value Rmax = M max max = 19 max. It was argued in Ref. 9, based on experiments in Ref. 3, that the alumina shell possesses maximum strength at least up to = 7.7 nm, i.e., Rmax = 146.3 nm. The increase in the strength of the oxide shell due to optimal passivation condition, creating a homogeneous shell thickness, and reducing inclusions of other elements should also lead to an increase in V for M 19. The substitution of the oxide shell with a stronger metal than alumina such as tungsten will promote the MDM and increase the reaction and ame velocity, in contrast to current trends.10 Tungsten also increases the density of the nanocomposite which is important for applications of reactive materials.

To summarize, the experimental data on ame velocity versus Al nanoparticle size and oxide shell thickness strongly support the MDM and contradict the classical diffusive oxidation mechanism. Namely, the main parameter controlling V is the ratio of the Al core radius to shell thickness M rather than R, , and Al content separately. Below the critical value M = 19, V reaches a maximum value and is independent of M, R, , and Al content. For M 19, V grows with reduction in M. In this case, an increase in oxide thickness under xed R leads to an increase in V. These results will signicantly change the directions in which Al nanoparticle synthesis progress. The current trends are directed at decreasing the particle size and oxide thickness while trying to keep the Al content high. The results suggest that the preferred course is to increase the oxide shell thickness up to a maximum value at which it is still possess the maximum strength due to a lack of defects, and to increase the particle size up to the value Rmax = M max max = 19 max. The increases in the oxidation temperature and strength of the shell an alumina shell or a specially designed shell made of a strong metal such as tungsten will increase V as well. It was found that V is proportional to the volume fraction of the molten Al that causes the fracture of the shell and participates in the MDM. This correlation suggests that in the given range of parameters cavitation of the molten Al does not limit the reaction. An analytical equation for V versus Al nanoparticle geometric parameters, thermomechanical properties, and synthesis parameters was derived. The maximum particle diameter corresponding to the fastest V was increased from 80 nm in Ref. 1 to 120 nm by increasing the oxide shell thickness to 4 nm. The authors gratefully acknowledge the ONR N0001407-1-0318 . M. Pantoya acknowledges the ARO W911NF04-1-0217 .
B. S. Bockmon, M. L. Pantoya, S. F. Son, B. W. Asay, and J. T. Mang, J. Appl. Phys. 98, 064903 2005 . 2 K. B. Plantier, M. L. Pantoya, and A. E. Gash, Combust. Flame 140, 299 2005 . 3 J. Granier and M. L. Pantoya, Combust. Flame 138, 373 2004 . 4 K. Park, D. Lee, A. Rai, D. Mukherjee, and M. Zachariah, J. Phys. Chem. B 109, 7290 2005 . 5 A. Rai, K. Park, L. Zhou, and M. Zachariah, Combust. Theory Modell. 10, 843 2006 . 6 V. Rosenband, Combust. Flame 137, 366 2004 . 7 M. A. Trunov, M. Schoenitz, X. Zhu, and E. L. Dreizin, Combust. Flame 140, 310 2005 . 8 V. I. Levitas, B. W. Asay, S. F. Son, and M. L. Pantoya, Appl. Phys. Lett. 89, 071909 2006 . 9 V. I. Levitas, B. W. Asay, S. F. Son, and M. L. Pantoya, J. Appl. Phys. 101, 083524 2007 . 10 Current trends are to substitute the oxide shell with metals such as gold, nickel, and silver T. J. Foley, C. E. Johnson, and K. T. Higa, Chem. Mater. 16, 4086 2005 or C13F27COOH R. J. Jouet, J. R. Carney, R. H. Granholm, H. W. Sandusky, and A. D. Warren, Mater. Sci. Technol. 22, 422 2006 to reduce the dead weight and initial barrier for diffusion. This is useful for some applications but not for those where the goal is to increase the reaction and ame velocity, because the strength of such shells is signicantly lower compared to alumina and MDM cannot operate. Indeed, even bare Al nanoparticles with D = 20 nm require more than 1 s for complete oxidation at T = 900 C Ref. 4 .
1

Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp